User interactions with items are driven by diverse intentions, and effectively modeling these intentions can greatly enhance recommendation systems’ efficacy. In this paper, we propose a Time-aware Intent Contrastive Learning (TICL) framework for sequential recommendation. Unlike existing methods, TICL employs contrastive learning in a hierarchical manner over augmented views, enabling robust sequence representation at each timestamp. The framework incorporates two contrastive tasks: temporal-aware and intent-wise contrastive learning. To address data sparsity across different time frames, two novel data augmentation operators, namely Down-Sample and Time-Warping are introduced. Additionally, the Rare-class Sample Generator (RSG) is presented to tackle the challenges posed by temporal data imbalance. By integrating temporal intentional data into the sequential recommendation framework, TICL significantly enhances recommendation accuracy and relevance. Empirical analysis on real-world datasets demonstrates that TICL outperforms existing baseline methods and exhibits resilience to data sparsity and interaction noise, leading to improved model performance where the maximum improvement can reach to 46.24%.